Arsenic is a common soil and groundwater contaminant in much of the world. Groundwater arsenic is commonly derived from the reductive dissolution of arsenic-bearing iron oxides. Despite considerable efforts, it is still difficult to predict the aqueous concentration of arsenic in the transitional redox environments where arsenic release occurs, or in the reduced sediments through which it is transported. Equilibrium-based partitioning models often no more accurate that simple adsorption isotherms or partition coefficients in these environments. Here, we develop and evaluate a novel kinetics-based approach that incorporates knowledge of the solid-phase, to predict arsenic concentrations. This model defines the steady-state aqueous concentration of arsenic as a function of iron oxide dissolution rate and readsorption of arsenite and arsenate. This model uses measured iron and arsenic redox status to successfully describe As concentrations in redox profiles in heterogeneous Bangladeshi and Vietnamese aquifers. These data imply that the speciation of arsenic, which affects readsorption rates, are the dominant variable controlling As retention, and that iron mineralogy plays an indirect role by affecting the rate of biological reduction.